Module 1: Incidence and prevalence Flashcards
PECOT
- Population: group of people who share a
specified common factor. - Exposure group
- Comparison group
- Outcomes:
• EGO: occurrence of dis-ease in exposure
group.
• CGO: occurrence of dis-ease in comparison
group. - An average can be taken and EG
compared to CG.
Incidence
- Incidence: counting the number of onsets of disease events occurring during a period of time.
• Longitudinal measure producing a rate.
• Most appropriate for observable events.
• Require dis-ease outcome to be categorical
variable.
• Measuring prevalence at two points of time
and calculating the change in prevalence bt/
the two points in time is a measure of the
incidence of dis-ease over the period bt/ the
two time points
ADVANTAGES
- Most useful for measuring causes of dis-ease
occurrence.
- Incidence is determined only by the dis-ease
risk in the population.
- Measures of incidence include events,
population and time.
DISADVANTAGES
- Incidence can be difficult to measure as one has
to observe events over time.
Prevalence
Prevalence: counting the number of people w/ a
dis-ease at a point in time.
• Cross sectional measure producing a figure.
• Most appropriate when transition from a nondis-eased state to a dis-eased state cannot
easily be observed and counted.
ADVANTAGES
- Prevalence is relatively easy to measure, as it is
static, taken for one point in time.
- Useful to funders and planners of health.
DISADVANTAGES
- Prevalence measures only include events and
population.
- Prevalence is determined by incidence, death
rate and cure rate
Prevalence
POINT PREVALENCE
- Point prevalence: outcome does not take any
previous time period into account and is simply
measured at one point in time.
PERIOD PREVALENCE
- Period prevalence: outcome/numerator depends
on the time period specified.
• Dis-ease outcomes cannot easily be
measured at one point in time, so we look
back and measure them over a period of
time.
A population could have a high incidence/low
prevalence if death/cure rate is also high.
A population could have a low incidence/high
prevalence if death/cure rate is also low.
C o m p a r i s o n s — R i s k
Differences and Relative Risk
- Differences bt/ EGO and CGO can provide
insight into the size of the effect of the study
exposure on the dis-ease outcome. - Comparisons of dis-ease occurrence typically
called ‘estimates of effect/association’ of an
exposure on a dis-ease outcome.
RISK DIFFERENCE
- EGO-CGO
- Units — same unit as EGO/CGO calculation e.g.
per x people over y years — more info. - Risk ratio can be the same whereas risk
difference can be much smaller/larger. - Also called difference in occurrences, absolute
risk (difference). - RD=0, no difference in effect of E and C on the
study outcome. - Risk difference is an absolute risk reduction if
risk is lower in the exposure group or an
absolute risk increase if the risk is higher in the
exposure group.
RELATIVE RISK
- EGO/CGO
- Also called risk ratio, relative risk difference,
ratio of occurrence. - No units — less information.
- RR=1, ‘no-effect’ value.
- If dis-ease occurrence measures are calculated
as averages, relative comparison of two mean
scores is ‘relative mean’ (RM). - Relative risk reduction: relative risk is subtracted
from 1.0, then expressed as a percentage. - Relative risk increase: 1.0 subtracted from
relative risk, then express as percentage.
Non Random Error
- Also called biases/systematic errors.
- If error occurs because of poor study design,
processes or measurement.
• Valid study: small amount of random/nonrandom error.
RECRUITMENT -RAMBOMAN
- Are participants representative sample from a
defined population? - Described as external validity error, as when
present findings may not be applicable to wider
population. - Particularly important when major objective of
study to measure characteristics of real population but participants recruited not
representative of eligibles.
Selection bias: participants allocated to EG
different source to participants allocated to CG.
• Confounding error caused by allocation
process.
- Non response bias: non-responders different to
responders. Consider response rate.
ALLOCATION
How well were participants allocated to EG and
CG?
- Confounders: EG and CG differ in ways other
than allocation, and these other difference have
an effect on the study outcome.
- RCTs: allocation by random process; all
participants have equal chance of allocation to
EG or CG, so groups are similar.
• RCTs are known as experiments, because
investigators actively control allocation
process.
• RCTs best way to stop confounding. - Complete baseline comparison to ensure
RCTs w/ small samples don’t have different
groups just through chance alone
Concealment of allocation stops tampering w/ randomisation process. - Observational studies: allocate by measurement and assign to EG and CG accordingly. • People may lie/under-report to hide embarrassment or they can’t remember. - Avoid by good questionnaire design. • Inaccurate measurement of exposure: (allocation) measurement error. • CG and EG often quite different: confounders if study outcome influenced. - Adjust for in analyses. - Sufficient information must be collected about other differences for adjustment purposes. - Confounding present in almost every observational study. • Two or more effects mix, all
MAINTENANCE
Will the validity of the study results be affected
by how well they were maintained in EG and
CG?
- Maintenance error: some participants’ exposure
status changes, or some are lost to follow-up.
• Did participants remain in their allocated
groups? Did they maintain their initial
exposure/comparison exposure?
- Long term cohort studies prone to maintenance
bias — offset by regular follow-ups.
- Maintenance not a problem for cross-sectional
studies.
BLIND AND OBJECTIVE MEASUREMENT
• Reduce error by blinding participants and
investigators to knowledge of which
intervention participants received.
• Blinding of outcome measurement ideal for
death certification when pathologist blind.
• R e d u c e e r r o r b y t a k i n g o b j e c t i v e
measurements where possible e.g. using a
machine.
- Measurement: use a standard definition.
ANALYSES
- Confounding can be reduced by dividing
participants into strata — stratified analysis e.g.
age standardisation.
Random Error
- Due to chance.
- No single study will ever measure the exact truth
in the whole population, even if it is a perfect
study.
• Every study an ‘estimate of the truth’.
• Identical studies will produce different results. - All measures of EGO/CGO/RR/RD/NNT have
random error.
• Most random errors can be reduced by
i n c re a s i n g s a m p l e s i z e , re p e a t i n g
measurements. - Extreme events are often chance events:
repeating measurements or studies w/ extreme
results many times usually gives less extreme
results — regression to the mean.
RANDOM SAMPLING ERROR
- Inherent in every study.
• N o s a m p l e w i l l e v e r b e p e r f e c t l y
representative of the population. - Every sample will differ due to chance, and
never include participants w/ identical
characteristics. - Bigger sample, smaller differences bt/ sample
and population
RANDOM MEASUREMENT/ASSESSMENT ERROR
- R a n d o m m e a s u r e m e n t e r r o r e ff e c t s
measurement of both exposures and outcomes. - Ability to measure biological factors the same
way every time is often poor, particularly when a
human operator is involved.
• Avoid by repeated measurements w/ an
average, or use, automatic, objective
machine. - Randomness inherent in biological phenomena.
• Inherent variability in biological phenomena,
and therefore in its measurement.
• Identical measurements of exposures and
outcomes in the same or similar people can
change from moment to moment.
RANDOM ALLOCATION ERROR
- Groups in an RCT may differ by chance alone,
particularly if trial is small. - Reduce by undertaking a larger study.
CONFIDENCE INTERVALS
- Every epidemiological measure has random
error in the estimate of the truth (EGO, CGO, RR,
RD) in the population that the study participants
were recruited from which can be estimated by a
confidence interval. - 95% CI definition: in 100 identical studies using
samples from the same population, 95/100 of
the 95% CIs will include the true value for the
population.
• There is about a 95% chance that the true
value in the population from which the
participants were recruited lies within the
95% CI, assuming no random error.
• Bigger the study, narrower the CI. - As number of events in study increases, recruited. Probably no statistically significant difference
• 95% CIs for RR and RD usually cross no effect line.
• Study results not statistically significant.
• If 95% CIs of RR and RD cross no effect line,
best stated as too much random error to
determine if there is a difference bt/ EGO and
CGO, as opposed to stating there is no
statistically significant difference. - 95% CI just touches no-effect line, study
“borderline statistically significant” (compare w/
statistically/not statistically significant). - Statistically significant event may be clinically
significant if a clinician would make a similar
decision whether the true result was near one
end of the confidence interval or the other.
Similarly, a small but statistically significant
effect w/ a narrow confidence interval may not
be of clinical significance.
width of 95% CIs decrease.
• Wider the interval, more random error in
measure.
• 95% CI most common for epidemiological
studies.
• 99% CI wider than a 95% CI.
• Confidence limit: each end of CI. - If no overlap of confidence intervals, reasonable
to assume that EGO and CGO are truly different
from each other in the underlying population.
• When there is no overlap of CIs in EGO and
CGO, confidence intervals for RD and RR will
not cross no-effect line.
• Therefore measures of association bt/ EGO
and CGO show real effect and ‘study results
are statistically significant’. - If overlap of CIs, study unable to determine if
EGO is different from CGO in the population
from which the study participants were
META ANALYSES
Meta-analysis: analytical technique use to
combine results of studies mathematically to
generate a summary estimate of the effect. The
result could be statistically significant.
- Next best thing to a large study and reduces
random error in effect estimates.
• Does not reduce non-random error.
- Most studies too small to produce precise
estimates (i.e. w/ low random error) of disease
frequency therefore measures of association
also generally imprecise.
- Point estimates for different studies all show a
RR difference, but the 95% CIs cross the no
effect line.
• Conventionally, none of the trials are
statistically significant.
• Does not necessarily mean treatment not
beneficial, as studies could simply have been
too small.
• Studies must have low levels of bias (nonrandom error) and reasonably similar findings.
• Combining multiple similar studies in a metaanalysis is an alternative to conducting one
large study.
- Meta-analyses commonly undertaken to
combine results of multiple RCTs which
individually have too much random error to
demonstrate whether or not intervention
has a real effect.
• Ideal meta-analyses combine studies from
systematic reviews, however meta-analyses
can be taken of a non-systematically
recruited group of studies, although one
should be wary of the latter.
Dahlgren and Whitehead and Causes of the causes
DETERMINANTS FOR INDIVIDUALS
- Any event, characteristic or other definable entity that brings about a change for better or worse
in health.
• Age, sex, ethnicity
• Water, shelter, sanitation (basic survival)
• Income, occupation, employment, education (provides opportunities for housing etc.)
• Housing, housing tenure, overcrowding, neighbourhoods (design of urban environments, is
one supportive and supported?, social and built environments)
• Access
• Deprivation
• Societal characteristics e.g. racism, attitudes to alcohol or violence, value on children (social
norms)
• Unconscious bias — racism, bullying, harassment
• Autonomy and empowerment — social cohesion (attitudes, resilience)
- What causes people to engage in activity which is a risk/protective factor — determinant of
determinant.
DETERMINANTS AND THE LIFE COURSE
- Determinants of health may vary at different life stages.
- Three ways in which life course events can interact to influence long-term health and wellbeing:
• Cumulative — notion of poverty trap, born into more or less wealth, children of parents of
grandparents.
• Multiplicative — multiplied effects on risk— multiple CVD risk factors combine to give
fractional increase.
• Programming — Barker hypothesis — genetics, in utero impacts critical periods in one’s
development.
DAHLGREN AND WHITEHEAD MODEL
Person — micro (individual) - Non-modifiable genes and hereditary, and individual lifestyle factors, e.g. attitudes to smoking, alcohol, diet, exercise, influenced by social and community networks.
• Community — meso (family, living, work)
- Local influences such as home, workplace and neighbourhood (behaviour of family,
friends, groups like me, community and its makeup.), and wider societal levels such as
health and education.
• Environment — macro (national/global)
- Cultural, social, political, physical and built
environment.
- Shared beliefs, values, attitudes and culture of broad
groupings.
- Govt regulation and policy.
Upstream and downstream
UPSTREAM VS DOWNSTREAM INTERVENTIONS
- Downstream interventions operate at a micro (proximal) level, including treatments systems,
and disease management.
- Upstream interventions operate at the macro (distal) level, such as government policies and
international trade agreements.
PROXIMAL AND DISTAL DETERMINANTS
Proximal determinants of health near (directly and readily associated e.g. lifestyle and
behavioural factors) the change in health status.
- Distal determinants of health are either distant in time or place from the change in health status.
• Also referred to as upstream factors, e.g. national, political, legal and cultural factors that
indirectly influence health by acting on proximal factors.
STRUCTURE AND AGENCY
- Structure: social and physical environmental conditions/patterns (social determinants) which
influence choices and available opportunities — upstream (three outer arches). - Agency (empowerment): capacity of an individual to act independently and make free choices
— downstream (two inner arches).
Māori health: HISTORY
- Early contact
• Māori initially flourished economically and
socially.
• Beginning of complex changes. - Official engagement
• Colonisation, Declaration of Independence,
Treaty of Waitangi, New Zealand.
• Herald of an era of depopulation, disease and
dispossession. - Colonisation
• Not value-free.
• Assumptions held by colonisers.
• Notions of superior and inferior peoples.
• Notions of civilisation especially religious but
also economic and scientific e.g. land use,
conservation.
• Notions of deserving and undeserving. - Societal barriers still obvious today.
IMPLICATIONS OF TREATY
Creation of Government — Article I • Construction of state sector — justice system, education, health, welfare etc. • Constitution Act 1852 created settler Government, determined voting rights - Laws and policies — Article II • Disregard for Māori voice/authority despite Article II - Māori land — historical basis of settler wealth • Pre-emption clause of Treaty - Māori land court — 1860s - Individual title - Different or denied citizenship — Article III • Old-age pensions 1898 - Equal provisions for Māori and pakeha. - Asian excluded. - Māori access difficult — through Māori land court. - Māori regularly removed from rolls. - Reduced amount paid to Māori. • Social Security Act 1938 - Underpayment continued after WWII.
RELATIONSHIP OF TREATY TO HEALTH
- Policy alienation.
- Land alienation:
• Social disruption of community.
• Breakdown of political power and alliances.
• Economic resource depletion and poverty.
• Resentment by indigenous peoples. - Unequal (inferior) citizenship:
• Entrenchment of poverty and dependance.
• Increased barriers to development.
• Acceptance of inequity by non-indigenous
groups.
• Resentment, frustration and anger.
• Social breakdown, crime, high risk behaviours.
Prevention, promotion, protection: approaches to
taking action
Population-based (mass) strategy
- Whole population.
- Reduce health-risks/improve outcome for all individuals of
population.
- Useful for common disease/widespread cause.
• Immunisations, seatbelts, low salt foods in supermarkets.
ADVANTAGES
- Radical, addressing underlying causes.
- Large potential benefit for whole population.
- Behaviourally appropriate — change social norms; something becomes acceptable.
DISADVANTAGES
- Small individual benefit.
- Poor individual motivation.
- Whole population exposed to downside of strategy (less favourable benefit to risk ratio).
High-risk individual strategy
- Focuses on perceived high-risk individuals.
- Well matched to individuals and their concerns.
• Obese adult interventions, intravenous drug users (NZ Needle Exchange), high systolic BP
individuals.
ADVANTAGES
- Tailored and targeted; appropriate to individuals.
- Individual motivation.
- Cost-effective use of resources.
- Favorable benefit to risk ratio.
DISADVANTAGES
- Cost of screening; need to identify individuals, and
- Temporary effect (e.g. breast screening cohort).
- Limited potential (small sample size).
- Behaviourally (culturally) inappropriate.
Health Promotion
Acts on determinants of wellbeing.
- Positive, health/wellbeing focus.
- Enables/empowers people to increase control over
— and improve — their health.
- Involves whole population in everyday contexts.
- Primary-care orientated.
OTTAWA CHARTER FOR HEALTH PROMOTION (WHO
21 November 1986.
- First International Conference on Health Promotion.
- ‘Mobilise action for community development’.
- Health is:
• A fundamental right for everybody.
• Requires individual and collective responsibility.
• Opportunity to have good health should be equally available.
• Good health is an essential element of social and economic
development.
- Three basic strategies:
• Enable — individual
- Provide opportunities for all individuals to make healthy choices through access to
information, life skills and supportive environments.
• Advocate — systems
- Create favourable political, economic, social, cultural and physical environments by
promoting/advocating for health and focusing on achieving equity in health.
• Mediate — individuals, groups and systems
- Facilitate/bring together individuals, groups and parties w/ opposing interests to work
together and compromise for health promotion.
- Five priority action areas:
• Develop personal skills: life skills education, ICT for health, awareness campaigns.
• Strengthen community action (community empowerment): Self-help groups and community
organised services (e.g. HIV/AIDS), initiatives to promote healthy schools, cities etc.
• Create supportive environments: air control, water and sanitation, speed bumps, work safety.
• Reorient health services towards primary health care: care responsive to needs of patients/
families (considering culture, social norms, physical and mental capacities, healthcare skills,
aspirations, resources), healthcare education, amenities to enhance hospital experience.
• Build healthy public policy: Tax alcohol and cigarettes, seat-belt use, ban smoking in public
places, food and drug control, mandatory sports in schoo
Disease Prevention
Disease focus.
- Looks at disease/injury and ways of preventing them.
- Primary: limit incidence by controlling specific causes and
risk factors e.g. seatbelts, immunisation.
- Secondary: reduce more serious consequences of disease e.g. hip fracture screening, lifeguard
services.
- Tertiary: reduce progress of complications of established disease: counselling services for
PTSD, rehabilitation for burns patients.
Health Protection
Predominantly environmental hazard focused.
- Risk/hazard assessment e.g. environmental
epidemiology, biosecurity, safe air and water.
- Monitoring e.g. biomarkers of exposure to
hazardous substances.
- Risk communication e.g. relating environmental
risks to public.
- Occupational health e.g. workplace safety
regulations.
Establishing causality in
population health
Association and causation
- Explore the concepts of association and causation in
understanding the determinants of health and
attempts to improve population health.
EPIDEMIOLOGY AND THE CAUSES OF DISEASE
- Epidemiology, causes of disease, appropriate
preventative measures can be introduced.
- Not determine disease cause in individual, but
relationship/association bt/ exposure—disease in
populations.
• Links exposures—outcomes.
• Lind’s controlled-trial of 12 sailors (small sample
size) showed citrus consumption associated w/
scurvy prevention.
- Preventative action taken before the cause
identified
Bradford Hill and Rothman’s Criteria
BRADFORD HILL CRITERIA (1965)
- Establish causality. - Aid, not definitive checklist.
- Not all criteria needed to establish causality.
• Temporality.
- Essential.
- First cause then disease.
• Questionnaire to establish exposures then
deaths followed up.
• Strength of association.
- Stronger association, causality more likely, in
absence of known biases (selection,
information, confounding).
• Consistency of association.
- Replication of findings by different investigators,
different times, places, different methods.
• Biological gradient (dose-response).
- Incremental change in disease rates—
corresponding changes in exposure.
• Biological plausibility of association.
- Does the association make sense biologically?
• Specificity of association.
- Weakest criteria. - A cause leads to a single effect, although a
single cause often leads to multiple effects.
• Reversibility.
- Demonstration that under controlled conditions,
changing exposure causes a change in the
outcome.
ROTHMAN’S CAUSAL PIES
- A cause of a disease: event, condition, characteristic
(or combination) playing essential role in producing
disease.
• Sufficient cause (causal mechanism): factor(s)
inevitably producing specific disease.
• Component cause: factor contributing to disease
causation, not sufficient alone to cause disease.
• Necessary cause: factor (component cause) must
be present for disease to occur. - Use association/other factors to infer causation and
intervene to prevent disease. - Can intervene at any point of pie.
- Knowledge of complete pathway not prerequisite for
introducing preventative measures. - Removing one of component causes can prevent
disease.
Screening: a special type
of prevention strategy
S u c c e s s f u l s c r e e n i n g
initiatives
SCREENING FUNDAMENTALS
- Screening: identifying risk factors for
disease/unrecognised disease by applying
tests, large scale to population.
- Can be primary (alcohol intake to prevent
breast cancer)/secondary (breast cancer
screening)/tertiary (screening for bone
density after breast cancer chemotherapy).
SCREENING COMPONENTS
COMPONENTS - Eligible population (most susceptible to disease). - Screening test: simple, cheap, easily applied to many. - Diagnostic test: best possible test for testing disease presence. - Intervention/treatment. - Re-screening at given time interval.
Diseases and screening tests
SCREENING CRITERIA - Suitable disease: • Important public health problem: relatively common/relatively uncommon but early detection and intervention leads to better outcome. • Knowledge natural history disease (or relationship risk factors—condition). - Detectable early (detectable risk factor/ disease marker). - Increased pre-clinical phase duration — screening more effective: more time to screen, treat. - Suitable test: • Reliable (consistent results), safe, simple, affordable, acceptable, accurate (test’s ability, indicate which individuals have disease/not). - Accuracy: gold standard diagnostic test, but can’t be applied to large populations. Screening test judged against diagnostic test, less expensive. - Suitable treatment: • Evidence early treatment->better outcomes. • Effective, acceptable, accessible treatment. • Evidence-based policies covering eligibles for treatment/appropriate treatment to be offered. - Suitable screening programme: • Benefits must outweigh harm. • RCT evidence screening programme will result in: - Reduced mortality. - Increased survival time. • Lead time bias. • Length time bias.
Screening test performance
DEFINITIONS
- True positive: disease, positive test.
- False negative: disease, negative test.
- False positive: no disease, positive test.
- True negative: no disease, negative test.
INTERPRETATIONS OF SENSITIVITY AND
SPECIFICITY
• Sensitivity high if proportion of true positives high. • Specificity high is proportion of true negatives is high. - Fixed characteristics of tests. • In two equally sized populations w/ different prevalences of disease, predictive values will be different.
Prioritizing in public health prioritization
EVIDENCE-BASED MEASURES - Descriptive: • Consider current problem (outcomes), who most/least affected (e.g. Treaty obligations may prioritise one disease over another, e.g. even though stroke overall greater death rate, respiratory disease rates greater for Māori
- Explanatory: • What are determinants/risks? Why getting worse/better? Why populations different? Epidemiological measures used in prioritisation e.g. determine major risk factors of disease burden.
- Evaluative: • What can improve health outcomes and in whom? How well can problem be solved? - Target population, expected number in population reached, evidence of effectiveness (based on known success rates), cost.
- Age at death/premature mortality — years of potential life lost to death (YLL). - Time lived w/ disability — years lived w/ disability (YLD). - Population Attributable Risk. • D o e s p r o b l e m / r i s k f a c t o r disproportionately affect population sub groups? Why? ToW considerations.
• Economic feasibility: economic sense to address problem/consequences of not doing so? • Opportunity cost: potential other health benefits had money been spent on next best alternative intervention/healthcare programme. • Acceptability: Will community/target population accept problem being addressed? Competing interests?
Population Attributable Risk
ATTRIBUTABLE RISK
- RD=AR (EGO-CGO).
• EGO = incidence in exposed population.
- Amount of extra disease attributable to
particular risk factor in exposed group.
POPULATION ATTRIBUTABLE RISK
- Amount of extra disease attributable to
particular risk factor in particular population.
- If association causal, amount of disease
theoretically preventable if particular risk
factor removed from population.
- RR: Compared w/ non-smoking mothers,
smoking mothers are four times more likely
to have a pre-term baby
- RD: Smoking mothers will have 600 more
preterm babies per 1000 women than nonsmoking mothers. - PAR: For every 1000 mothers in population,
we can prevent 90 preterm births by
removing heroin addiction as a risk factor. - PAR higher when disease prevalence higher
in population.
